Simulations of drifting fish aggregating device (dFAD) trajectories in the Atlantic and Indian Oceans

Tropical tuna purse-seine fisheries deploy thousands of human-made drifting fish aggregating devices (dFADs) annually, raising a number of concerns regarding ecosystem impacts. In this study, we explored the use of a Lagrangian particle-tracking model to simulate the drift of dFADs in the Atlantic and Indian Oceans. We simulated more than 100,000 dFADs trajectories using the Lagrangian tool Ichthyop forced with velocity fields from an ocean model output (GLORYS12V1) and two satellite-derived ocean currents products (OSCAR and GEKCO). Importantly, through a collaborative agreement with the French frozen tuna producers' organization we had access to the true locations of all dFADs along their drift and could therefore evaluate the accuracy of our simulations. The accuracy was assessed by comparing the observed and simulated trajectories in terms of spatial distribution, separation distance, and a non-dimensional skill score (an index based on separation distances normalized by net displacements of dFADs). In the two oceans, simulations forced with GLORYS12V1 were more accurate than with OSCAR and GEKCO, probably due to the differences in the spatio-temporal resolution of the forcing products. When we compared multiple depths for GLORYS12V1, the model performed better at 0 m in the Indian Ocean and at 5 m in the Atlantic Ocean, which could be related to the longer vertical structure of dFADs in the Atlantic Ocean. We showed that including a windage factor did not improve the accuracy of modeled dFADs trajectories. We found that mean model-data separation distances were similar in both oceans, exceeding 100 km after 6-8 days of drift. While separation distances between simulated and observed trajectories show that model errors were similar in the two oceans, the generally longer distances traveled by dFADs in the Indian Ocean than in the Atlantic Ocean lead to considerably higher skill scores in the former than in the latter. This explains the relatively good predictive ability of the model to represent mean dFAD densities at the basin scale in both oceans, while at the same time indicates higher prediction skills for the movements of individual dFADs in the Indian Ocean than in the Atlantic Ocean.

Automated summary

This study used a Lagrangian particle-tracking model to simulate the drift of drifting fish aggregating devices (dFADs) in the Atlantic and Indian Oceans. The simulations showed that using the ocean model output (GLORYS12V1) provided more accurate results compared to satellite-derived ocean currents products (OSCAR and GEKCO). The study also found that the predictive ability of the model was higher for individual dFAD movements in the Indian Ocean than in the Atlantic Ocean.